Defining element associations and inferring geological processes from total element concentrations in Australian catchment outlet sediments: Multivariate analysis of continental-scale geochemical data

In this paper, the geochemical composition of surficial regolith is statistically analysed and compared to independent geoscientific datasets to infer processes governing regolith composition. Surface (0–10 cm depth) and sub-surface (∼60–80 cm depth) transported sediment samples from the National Geochemical Survey of Australia were analysed for total element content in both coarse (<2 mm) and fine (<75 μm) grain-size fractions. Multi-element total content data was obtained from mainly XRF and total digestion ICP-MS analysis, of which the 50 elements satisfying data quality criteria, plus Loss on Ignition, are used herein. ed data (<lower limit of detection) was imputed using a nearest neighbour-based analysis. The compositional data was transformed using centered log ratios (clr) to circumvent closure issues. A Principal Component Analysis (PCA) was then performed on the dataset. The first four PCs account for 59% of the variance in the dataset. Both negative and positive loadings of each of these PCs relate to geological processes consistent with the element associations they represent as well as the spatial distribution patterns they produce. The positive loadings of PC1 represent the accumulation of resistant minerals rich in Rare Earth Elements (REEs) that results from intense weathering, except in southeastern Australia where they reflect REE-enriched igneous and sedimentary rocks. Negative PC1 loadings represent secondary minerals formed during weathering (carbonates, sulfates, Fe-oxyhydroxides). Negative PC2 loadings are a mixture of elements (e.g., Co, Mn, Zn, V) characterising mafic and ultramafic minerals; conversely negative PC3 loadings (e.g., K, Rb, Na, Sr, Ca) represent more felsic minerals. Spatial distributions of the PCs are compared with independent spatial information from geological maps, airborne radiometric and spaceborne spectroscopic datasets. The differences between surface and sub-surface and between coarse and fine grain-size fractions are analysed. The implied processes (e.g., lithological control, weathering, transport, secondary mineral precipitation) overall match well with this new geochemical evidence. Future work directions with this dataset include lithological prediction and mineral prospectivity analysis.